diff --git a/src/crawlee/_utils/sitemap.py b/src/crawlee/_utils/sitemap.py
index 190381c1f2..c3690e2a50 100644
--- a/src/crawlee/_utils/sitemap.py
+++ b/src/crawlee/_utils/sitemap.py
@@ -37,6 +37,15 @@
 SITEMAP_URL_PATTERN = re.compile(r'\/sitemap\.(?:xml|txt)(?:\.gz)?$', re.IGNORECASE)
 COMMON_SITEMAP_PATHS = ['/sitemap.xml', '/sitemap.txt', '/sitemap_index.xml']
 
+MAX_SITEMAP_SIZE = 50 * 1024 * 1024
+"""Maximum sitemap size in bytes after decompression, per the sitemap protocol; content beyond it is truncated."""
+
+CONTENT_CHUNK_SIZE = 1024 * 1024
+"""Maximum size of a single decompressed content piece fed to the parser, keeping memory usage flat."""
+
+DEFAULT_MAX_DEPTH = 10
+"""Default maximum depth of nested sitemaps to follow, guarding against malicious infinite sitemap chains."""
+
 
 @dataclass()
 class SitemapUrl:
@@ -368,6 +377,9 @@ async def _fetch_and_process_sitemap(
                 # Create appropriate parser
                 parser = _get_parser(content_type, sitemap_url)
                 decompressor = None
+                content_bytes = 0
+                raw_bytes = 0
+                truncated = False
                 try:
                     # Process chunks as they arrive
                     first_chunk = True
@@ -377,20 +389,50 @@ async def _fetch_and_process_sitemap(
                             decompressor = zlib.decompressobj(zlib.MAX_WBITS | 16)
                         first_chunk = False
 
-                        chunk = decompressor.decompress(raw_chunk) if decompressor else raw_chunk
-                        text_chunk = decoder.decode(chunk)
-                        async for item in parser.process_chunk(text_chunk):
-                            async for result in _process_sitemap_item(
-                                item,
-                                source,
-                                depth,
-                                visited_sitemap_urls,
-                                sources,
-                                emit_nested_sitemaps=emit_nested_sitemaps,
-                                enqueue_strategy=enqueue_strategy,
-                            ):
-                                if result:
-                                    yield result
+                        raw_bytes += len(raw_chunk)
+
+                        # Decompress and parse in bounded pieces, capping the total decompressed size,
+                        # so that e.g. a gzip bomb cannot exhaust memory. Slice via a memoryview to avoid
+                        # repeatedly copying the tail when a single chunk is much larger than a piece.
+                        pending: memoryview | bytes = memoryview(raw_chunk)
+                        while pending and content_bytes < MAX_SITEMAP_SIZE:
+                            max_length = min(MAX_SITEMAP_SIZE - content_bytes, CONTENT_CHUNK_SIZE)
+                            if decompressor:
+                                chunk = decompressor.decompress(pending, max_length)
+                                pending = decompressor.unconsumed_tail
+                            else:
+                                chunk, pending = pending[:max_length], pending[max_length:]
+                            content_bytes += len(chunk)
+
+                            text_chunk = decoder.decode(chunk)
+                            async for item in parser.process_chunk(text_chunk):
+                                async for result in _process_sitemap_item(
+                                    item,
+                                    source,
+                                    depth,
+                                    visited_sitemap_urls,
+                                    sources,
+                                    emit_nested_sitemaps=emit_nested_sitemaps,
+                                    enqueue_strategy=enqueue_strategy,
+                                ):
+                                    if result:
+                                        yield result
+
+                        # Stop once the decompressed-size cap leaves unprocessed bytes, or the compressed input
+                        # exceeds the cap (e.g. a stream that produces little or no output). Both are truncations.
+                        if pending or raw_bytes > MAX_SITEMAP_SIZE:
+                            truncated = True
+                            break
+                        # Stop once the gzip stream has ended; further bytes would only accumulate in the
+                        # decompressor's `unused_data` (trailing garbage or extra members are not processed).
+                        if decompressor and decompressor.eof:
+                            break
+
+                    if truncated:
+                        logger.warning(
+                            f'Sitemap {sitemap_url} exceeded the maximum size of {MAX_SITEMAP_SIZE} bytes, '
+                            f'processing only the truncated content.'
+                        )
 
                     # Process any remaining content
                     async for item in parser.flush():
@@ -472,7 +514,7 @@ async def parse_sitemap(
 
     This function coordinates the process of fetching and parsing sitemaps,
     handling both URL-based and raw content sources. It follows nested sitemaps
-    up to the specified maximum depth.
+    up to the specified maximum depth (`ParseSitemapOptions.max_depth`, 10 by default).
 
     Default `ParseSitemapOptions.enqueue_strategy` is `same-hostname` which will skip cross-host URLs.
     Use strategy `all` to process all links.
@@ -480,7 +522,7 @@ async def parse_sitemap(
     # Set default options
     options = options or {}
     emit_nested_sitemaps = options.get('emit_nested_sitemaps', False)
-    max_depth = options.get('max_depth', float('inf'))
+    max_depth = options.get('max_depth', DEFAULT_MAX_DEPTH)
     sitemap_retries = options.get('sitemap_retries', 3)
     timeout = options.get('timeout', timedelta(seconds=30))
     enqueue_strategy = options.get('enqueue_strategy', 'same-hostname')
diff --git a/src/crawlee/request_loaders/_sitemap_request_loader.py b/src/crawlee/request_loaders/_sitemap_request_loader.py
index 368347e708..47b026bbd4 100644
--- a/src/crawlee/request_loaders/_sitemap_request_loader.py
+++ b/src/crawlee/request_loaders/_sitemap_request_loader.py
@@ -14,7 +14,14 @@
 from crawlee._utils.docs import docs_group
 from crawlee._utils.globs import Glob
 from crawlee._utils.recoverable_state import RecoverableState
-from crawlee._utils.sitemap import NestedSitemap, ParseSitemapOptions, SitemapSource, SitemapUrl, parse_sitemap
+from crawlee._utils.sitemap import (
+    DEFAULT_MAX_DEPTH,
+    NestedSitemap,
+    ParseSitemapOptions,
+    SitemapSource,
+    SitemapUrl,
+    parse_sitemap,
+)
 from crawlee._utils.urls import filter_url
 from crawlee.request_loaders._request_loader import RequestLoader
 
@@ -80,6 +87,9 @@ class SitemapRequestLoaderState(BaseModel):
     processed_sitemap_urls: Annotated[set[str], Field(alias='processedSitemapUrls')] = set()
     """Set of processed sitemap URLs."""
 
+    sitemap_depths: Annotated[dict[str, int], Field(alias='sitemapDepths')] = {}
+    """Nesting depth of each known sitemap URL, used to bound how far nested sitemap chains are followed."""
+
     completed: Annotated[bool, Field(alias='sitemapCompleted')] = False
     """Whether all sitemaps have been fully processed."""
 
@@ -286,6 +296,8 @@ async def _get_state(self) -> SitemapRequestLoaderState:
             )
             if not has_sitemap_for_processing and not self._state.current_value.completed:
                 self._state.current_value.pending_sitemap_urls.extend(self._sitemap_urls)
+                for sitemap_url in self._sitemap_urls:
+                    self._state.current_value.sitemap_depths.setdefault(sitemap_url, 0)
 
             if self._state.current_value.in_progress:
                 self._state.current_value.url_queue.extendleft(self._state.current_value.in_progress)
@@ -352,6 +364,10 @@ async def _load_sitemaps(self) -> None:
                         continue
                     state.in_progress_sitemap_url = sitemap_url
 
+                # Depth of the sitemap currently being processed; nested sitemaps found in it are one level deeper.
+                # This bounds a malicious chain of unique nested sitemap URLs, which exact-URL dedup alone cannot.
+                current_depth = state.sitemap_depths.get(sitemap_url, 0)
+
                 parse_options = ParseSitemapOptions(
                     max_depth=0,
                     emit_nested_sitemaps=True,
@@ -375,9 +391,15 @@ async def _load_sitemaps(self) -> None:
                                 item.loc not in state.pending_sitemap_urls
                                 and item.loc not in state.processed_sitemap_urls
                             ):
+                                if current_depth >= DEFAULT_MAX_DEPTH:
+                                    logger.warning(
+                                        f'Skipping nested sitemap {item.loc!r}: max depth {DEFAULT_MAX_DEPTH} reached.'
+                                    )
+                                    continue
                                 if not self._passes_filters(item.loc, parsed_sitemap_url, 'nested sitemap'):
                                     continue
                                 state.pending_sitemap_urls.append(item.loc)
+                                state.sitemap_depths[item.loc] = current_depth + 1
                             continue
 
                         if isinstance(item, SitemapUrl):
@@ -413,6 +435,7 @@ async def _load_sitemaps(self) -> None:
                 state.in_progress_sitemap_url = None
                 if current_sitemap_url:
                     state.processed_sitemap_urls.add(current_sitemap_url)
+                    state.sitemap_depths.pop(current_sitemap_url, None)
                 state.current_sitemap_processed_urls.clear()
 
             # Mark as completed after processing all sitemap urls
diff --git a/tests/unit/_utils/test_sitemap.py b/tests/unit/_utils/test_sitemap.py
index e746c314c7..d209a4a3d4 100644
--- a/tests/unit/_utils/test_sitemap.py
+++ b/tests/unit/_utils/test_sitemap.py
@@ -10,6 +10,7 @@
 from yarl import URL
 
 from crawlee._utils.sitemap import (
+    DEFAULT_MAX_DEPTH,
     ParseSitemapOptions,
     Sitemap,
     SitemapUrl,
@@ -22,7 +23,7 @@
 from tests.unit.utils import DEFAULT_URL, get_basic_results, get_basic_sitemap
 
 if TYPE_CHECKING:
-    from collections.abc import AsyncIterator
+    from collections.abc import AsyncIterator, Callable
 
 
 def _make_mock_client(url_map: dict[str, tuple[int, bytes]]) -> AsyncMock:
@@ -66,6 +67,27 @@ async def read_stream() -> 'AsyncIterator[bytes]':
     return client, attempts
 
 
+def _make_stream_client(body_for_url: 'Callable[[str], bytes]') -> tuple[AsyncMock, list[str]]:
+    """Create a mock client whose `stream` serves `body_for_url(url)` in a single chunk and records fetched URLs."""
+    fetched: list[str] = []
+
+    @asynccontextmanager
+    async def stream(url: str, **_kwargs: Any) -> 'AsyncIterator[HttpResponse]':
+        fetched.append(url)
+
+        async def read_stream() -> 'AsyncIterator[bytes]':
+            yield body_for_url(url)
+
+        response = MagicMock(spec=HttpResponse)
+        response.headers = {'content-type': 'application/xml; charset=utf-8'}
+        response.read_stream = read_stream
+        yield cast('HttpResponse', response)
+
+    client = AsyncMock(spec=HttpClient)
+    client.stream = stream
+    return client, fetched
+
+
 def compress_gzip(data: str) -> bytes:
     """Compress a string using gzip."""
     return gzip.compress(data.encode())
@@ -317,6 +339,68 @@ async def test_sitemap_fetch_raises_after_retries_exhausted() -> None:
     assert len(attempts) == 3
 
 
+async def test_gzip_bomb_sitemap_truncated_at_size_cap(monkeypatch: pytest.MonkeyPatch) -> None:
+    """A gzip sitemap inflating past the size cap is truncated instead of being decompressed without bound."""
+    monkeypatch.setattr('crawlee._utils.sitemap.MAX_SITEMAP_SIZE', 64 * 1024)
+    total_urls = 20_000
+    locs = ''.join(f'<url><loc>{DEFAULT_URL}page-{i}</loc></url>' for i in range(total_urls))
+    # ~1 MB of XML arriving as a single small compressed chunk.
+    bomb = compress_gzip(f'<urlset>{locs}</urlset>')
+
+    client, _ = _make_stream_client(lambda _url: bomb)
+    items = [item async for item in parse_sitemap([{'type': 'url', 'url': f'{DEFAULT_URL}sitemap.xml'}], client)]
+
+    assert 0 < len(items) < total_urls
+
+
+async def test_gzip_sitemap_stops_reading_after_member_end() -> None:
+    """Trailing bytes after a complete gzip member must not be read; otherwise they accumulate unbounded."""
+    member = compress_gzip(get_basic_sitemap())
+    chunks_read = 0
+
+    @asynccontextmanager
+    async def stream(_url: str, **_kwargs: Any) -> 'AsyncIterator[HttpResponse]':
+        async def read_stream() -> 'AsyncIterator[bytes]':
+            nonlocal chunks_read
+            chunks_read += 1
+            yield member
+            # A malicious server would stream trailing junk forever; reading any of it is the bug.
+            while True:
+                chunks_read += 1
+                yield b'\x00' * 65536
+
+        response = MagicMock(spec=HttpResponse)
+        response.headers = {'content-type': 'application/gzip'}
+        response.read_stream = read_stream
+        yield cast('HttpResponse', response)
+
+    client = AsyncMock(spec=HttpClient)
+    client.stream = stream
+
+    items = [item async for item in parse_sitemap([{'type': 'url', 'url': f'{DEFAULT_URL}sitemap.xml.gz'}], client)]
+
+    assert {item.loc for item in items} == get_basic_results()
+    # The member fits in the first chunk, so at most the first trailing chunk may be pulled before we stop.
+    assert chunks_read <= 2
+
+
+async def test_nested_sitemaps_followed_only_to_default_max_depth() -> None:
+    """A chain of unique nested sitemap URLs is followed only up to the default max depth."""
+    chain_length = 3 * DEFAULT_MAX_DEPTH
+
+    def body_for_url(url: str) -> bytes:
+        index = int(url.removesuffix('.xml').rsplit('-', 1)[-1])
+        if index >= chain_length:
+            return f'<urlset><url><loc>{DEFAULT_URL}page</loc></url></urlset>'.encode()
+        next_sitemap = f'{DEFAULT_URL}sitemap-{index + 1}.xml'
+        return f'<sitemapindex><sitemap><loc>{next_sitemap}</loc></sitemap></sitemapindex>'.encode()
+
+    client, fetched = _make_stream_client(body_for_url)
+    _ = [item async for item in parse_sitemap([{'type': 'url', 'url': f'{DEFAULT_URL}sitemap-0.xml'}], client)]
+
+    assert len(fetched) == DEFAULT_MAX_DEPTH + 1
+
+
 async def test_parse_sitemap_with_partial_options() -> None:
     """Test that missing keys in partial `ParseSitemapOptions` fall back to defaults."""
     options = ParseSitemapOptions(timeout=timedelta(seconds=10))
diff --git a/tests/unit/request_loaders/test_sitemap_request_loader.py b/tests/unit/request_loaders/test_sitemap_request_loader.py
index abfd83cd62..a70c133afa 100644
--- a/tests/unit/request_loaders/test_sitemap_request_loader.py
+++ b/tests/unit/request_loaders/test_sitemap_request_loader.py
@@ -1,17 +1,21 @@
 import base64
 import gzip
-from typing import TYPE_CHECKING
-from unittest.mock import patch
+from contextlib import asynccontextmanager
+from typing import TYPE_CHECKING, Any, cast
+from unittest.mock import AsyncMock, MagicMock, patch
 
 from yarl import URL
 
 from crawlee import RequestOptions, RequestTransformAction
-from crawlee.http_clients._base import HttpClient
+from crawlee._utils.sitemap import DEFAULT_MAX_DEPTH
+from crawlee.http_clients._base import HttpClient, HttpResponse
 from crawlee.request_loaders._sitemap_request_loader import SitemapRequestLoader
 from crawlee.storages import KeyValueStore
-from tests.unit.utils import get_basic_results, get_basic_sitemap, poll_until_condition
+from tests.unit.utils import DEFAULT_URL, get_basic_results, get_basic_sitemap, poll_until_condition
 
 if TYPE_CHECKING:
+    from collections.abc import AsyncIterator
+
     from crawlee._types import JsonSerializable
 
 
@@ -25,6 +29,36 @@ def encode_base64(data: bytes) -> str:
     return base64.b64encode(data).decode('utf-8')
 
 
+async def test_nested_sitemap_chain_bounded_by_max_depth() -> None:
+    """A malicious endless chain of unique nested sitemaps is followed only up to the default max depth."""
+    fetched: list[str] = []
+
+    @asynccontextmanager
+    async def stream(url: str, **_kwargs: Any) -> 'AsyncIterator[HttpResponse]':
+        fetched.append(url)
+        index = int(url.removesuffix('.xml').rsplit('-', 1)[-1])
+        # Every sitemap is an index pointing at the next unique URL, so only the depth cap can stop the chain.
+        next_sitemap = f'{DEFAULT_URL}sitemap-{index + 1}.xml'
+        body = f'<sitemapindex><sitemap><loc>{next_sitemap}</loc></sitemap></sitemapindex>'.encode()
+
+        async def read_stream() -> 'AsyncIterator[bytes]':
+            yield body
+
+        response = MagicMock(spec=HttpResponse)
+        response.status_code = 200
+        response.headers = {'content-type': 'application/xml; charset=utf-8'}
+        response.read_stream = read_stream
+        yield cast('HttpResponse', response)
+
+    client = AsyncMock(spec=HttpClient)
+    client.stream = stream
+
+    loader = SitemapRequestLoader([f'{DEFAULT_URL}sitemap-0.xml'], http_client=client)
+    assert await poll_until_condition(loader.is_finished)
+
+    assert len(fetched) == DEFAULT_MAX_DEPTH + 1
+
+
 async def test_sitemap_traversal(server_url: URL, http_client: HttpClient) -> None:
     sitemap_url = (server_url / 'sitemap.xml').with_query(
         base64=encode_base64(get_basic_sitemap(url=server_url).encode())